MusiCat System Makes Library Searches More Fruitful THIS MONTH'S THEME:

This innovative 'search tree' design allows patrons to get to previously hard-to-reach information, because the design lets users think like - users!

A carefully constructed Web shell is allowing patrons to search one library catalog in a whole new way. Since the design eliminates the need for keywords and Boolean expertise, patrons can think like patrons and still get great search results. Read more in Computers in Libraries and take a look at The Music Treasure MusiCat.
Computers in Libraries, October 2002

MusiCat System Makes Library Searches More Fruitful
By Per Hofman Hansen, Silkeborg Public Library in Denmark
Published in Computers in Libraries, October 2002, Volume 22, Number 9, Pages 26-31.

Traditional library catalogs have been limited in the ways they allow library patrons to search and identify specific types of musical resources. MusiCat is a Web system that aims to empower users and to make library resources more accesible.

For a long time I have been aware of a paradox: Commercial interests such as Yahoo!, the Google Directory (started as the DMOZ Open Directory Project), and similar Internet directories have for years constructed indexes aimed at organizing information found on the Internet, but, as far as I know, no libraries have attempted to use a similar approach to organize their catalogs by "unfolding" and "visualizing" their content for their patrons. Traditional library catalogs have been restricted in the manner in which they allow library patrons to search and identify specific types of musical resources. Patrons have been left on their own facing an empty search field with a blinking cursor that seems to say: "Please be as clever and intelligent as the librarian who cataloged the records in the database!"

And so I have created MusiCat, an experimental user interface that reflects the different ways that patrons think when searching for musical material. MusiCat is a Web shell that aims to empower users to make the musical resources of the library more readily accessible, so that user searches are more fruitful.

Technology vs. Content: Which Takes Precedence?
Some people would probably call me a Webmaster, but as a librarian who designs and is in charge of the contents of our library's Web site, I prefer to call myself a Web editor, just like an editor of a newspaper. And why is that? Because I do not attach importance to the technical side of the job but much more to the content aspect - just as the librarian, the journalist, and the author do.

I work in the city of Silkeborg, Denmark, which has 52,000 inhabitants. The Silkeborg Public Library was the first Danish public Library with its own Web site (established in 1995). Since 1999, I have been the site's editor in chief. In 2001, our site bested 1,700 participants and won the gold medal in the Danish competition "Top of the Web" as the best public sector Web site in Denmark. I'm sure that one of the main reasons we won the gold medal was that since the beginning we have been thinking about how to best communicate with the patrons - placing our patrons' wishes and needs in focus rather than making complicated technical solutions that are not userfriendly. The following story has a lot to do with these thoughts and ideas. This is how I tried to "redefine the library experience" for my users.

Why We Need Trees
I am sure you have experienced the following situation: A patron has just asked you a complex question and you turn to the "oracle" of our time, the computer. An impatiently blinking cursor attempts to coax you to write some creative words. But how did the cataloger cllaify the material about the subject? What key words did she add to the records? How did she spell them? Were they in English? Were they singular or plural? You give it your best try and if the database is huge you will most likely get an answer. But what kind of an answer will it be? Are you always certain that you have gotten the best answer? How can you be sure that you didn't miss other valuable information? Most of the time you cannot be sure. Patrons are even less sure when they search for themselves.

As a Danish librarian with nearly 30 years of professional experience, I have worked with both the traditional card file catalog and electronic databases. We primarily introduced computer technology in the libraries for the purpose of automatation and economizing the daily routines, but we forgot our patrons. We threw out the old card files - which many librarians and patrons treasured - and replaced them with some primitive DOS- or UNIX-based yellow-brown terminals. At that moment the difficult task of making humans, computers, and interfaces work together to create good usability began!

My first attempt to help out our patrons in that situation was to put up a simple poster next to the computer to give them a few helpful hints. Since I am not a computer programmer I could not change the layout or the functionality on the monitors. I was obliged to find new ways myself because the library automation system companies did not hear our cries for help. The next logical step was to make a Windows Help file and install it on a separate computer. But it was not ideal. In 1993, we developed a finger-touch Windows-based system, which performed so well that we were able to convince our library automation system company to buy it.

Almost before we realized it, the Internet was here. It brought a new world, a new way of thinking, and also new ways to reach current and potential library users. They were located "somewhere out there in cyberspace," so the problems concerning usability have grown in some cases. The Internet is increasingly inviting its users to self-service. We love it: home banking, reading the gas meter, paying the electricity bill, etc., and of course accessing the library with its huge databases. The patron can control his record, renew books, place and review holds, check for fines and/or fees, find lists of new materials, receive news about the library by mail, etc. So we threw ourselves into using the resources on the Internet but forgot that the most precious information lay just at our feet - the library database with thousands or millions of authors, titles, annotations, content descriptions, and keywords. Why don't we make the best possible use of it?

"Technically, MusiCat is constructed
as a Web shell that lies over the music section
of our library catalog database."

Due to the different systematic entries of the commercial directories, the world of the Internet becomes visible all at once. Even the more-distant, hard-to-reach data can be discovered with relative ease because the information is organized in categories that make sense to everyday people. Order has been created from chaos, related subjects have been categorized together in a hierarchy. I visualize it like a tree, where general subjects (such as history or literature) make up the trunk of the hierarchy, the more specific subjects (such as American History or Irish Literature) make up the branches, then the most-specialized topics (such as the assassination of Martin Luther King, Jr. or James Joyce's Ulysses) form the leaves of the tree.

Because of this sort of hierarchy, this unfolding and visualization of topics and their relationships to each other, a Web user is presented with immense possibilities to find inspiration, form associations, and to move from branch to branch. However, in library catalog, this is not necessarily the case.

Why Aren't Librarians Designing Search Trees?
Why were the librarians not the first ones to use this obvious idea, even when the hypertext technology made it possible? I have often wondered about this. For the last 3 years I have conducted several experiments with the intention of revealing parts of the library database via the Web. However, the most ambitious project has been the attempt to open and visualize the world of music - in this case that means Silkeborg's collection of music scores, CDs, cassette tapes, and the old but often-valuable LP records found in the music section of the library.

This project gave me the idea of seriously throwing myself into a (nearly) Sisyphean work - making a detailed directory of musical resources. One of my fundamental ideas is that the patrons should not have to worry about classifications and call numbers. If it were possible to group and visualize the subject headings in a logical way, we could get more use out of some of the costly resources listed in the library database. But what was most important to me was to help colleagues and patrons to easily find what they're looking for, and to inspire them to look into neighbouring genres and subjects based on the associations they make by visualization.

Planning the First Tree
Technically, MusiCat is constructed as a Web shell that lies over the music section of our library catalog database. Musicat builts on the same idea or concept as Google's directory: It starts with a solid tree trunk, adds thick branches, then thinner branches, and at last all the green leaves. Here's one example from MusiCat: One main directory is jazz, the first subdirectory is traditional jazz, the next subdirectory is instruments, and at last comes the saxophone (the leaf). This is what I mean by unfolding the library database by visualizing the content and relationships between the records. Basically, the idea was to think like a searcher, not a cataloger.

But before beginning this great and complicated project I formed an alliance with my colleague, a professional problem solver and a specialist in music library work, Maja Frang. Experience told us that it was practical to divide the music into seven main groups: classical, rock, jazz, blues, folk music, easy listening, and miscellaneous. Each of the seven groups was in turn divided into a varying number of subgroups depending upon the size and number of genres.

In an analysis of the relevant subject headings connected to a defined Dewey group, in this case "Traditional jazz until 1935", we found the following subject headings:

Jazz (in general) 46
Traditional jazz 42  
Instruments 34  
USA 31  
Vocal 30  
1990-1999 23  
Denmark 16  
Swing 15  
Trumpet 12  
1940-1949 10  
Clarinet 11  
England 7  
Dixieland 6  
Piano 6  
1920-1929 5  
1930-1939 5  
1980-1989 5  
Saxophone 5  
Ragtime 4  
France 3  
Boogie Woogie 2  
Blues 2  
Sweden 1  
Charleston 1  

The list shows the number of records in the database where the subject headings were found. (The analysis resulted in 40 different subject headings, but only a selection is shown here.)

MusiCat is meant to be a real, practical, do-it-yourself tool that should be easy for anyone to use. Trying to organize things and thinking practically, we decided to neglect subject headings if they covered less than four or five records in the database. On the other hand, if we were able to gather neighboring but seldom-used subject headings together and place them under a single heading, we would be doing our best. Actually, this means that in the list of jazz headings, we omitted the meaningless subject heading "jazz", whereas instead we defined three obvious subgroups for instruments, periods, and countries respectively. Saxophone and tenor saxophone are gathered under one single heading.

During the past year, Maja and I met for approximately 2 hours every week, during which time we constructed MusiCat while discussing a number of different outlines and drafts. We discussed topics like the relationship between subjects, the use of controlled library terms vs. popular slang, how to construct the best search string to get a not-too-broad and not-too-narrow result, how to spell and transcribe Japanese or Chinese words for different instruments, and so on. For example: The Danish edition of the Dewey Decimal Classification system uses the subject heading "evergreens" when talking about classic jazz, but jazz enthusiasts use the term "standards". Which should we use? We had no doubts: We chose the term "standards", keeping the end-user in mind.

If you take a look at the Danish edition of Dewey you will find that jazz is divided into seven groups nearly equal to the main groups found in the Jazz section of MusiCat. MusiCat uses the seven groups as the basis for more than 230 links to subgroups or subject headings, beneath which lie hidden search queries constructed as combinations of call numbers and subject headings.

Cultivating More Trees
When we looked at the following sections - hip hop, rap, reggae, techno, and world music - the Danish edition of Dewey showed even greater deficiencies. These new genres cannot be found by using call numbers, but must be searched for exclusively as free-text or with keywords or subject headings.

Here's a screen from MusiCat with European folk music grouped by countries. Let's say you want to listen to some traditional dance and folk music from Sweden called polka. But how can you find it? With our 'tree' design, it's easy! Move to Folk Music-Europe, then jump to Sweden and select Polka (the link is red). Click the link and you get eight records in the database with Swedish Polka Folk Music. (At the top you can choose between four types of media: CDs, cassette tapes, records, and music books).
With the intention of forming an overview, we analysed the distribution of subject headings. Our next step was to group the subject headings into a number of main groups. Many bibliographical records are identified by keywords and they often define the basic structure. After only a few investigations we noticed that a lot of interesting genres were missing. We often found that the database contained several more relevant bibliographical records than we were able to find by using keywords. To improve the search success rate, we have in many cases added a free-text search to the query. Fortunately, many terms are often so specific - boogie woogie, bossa nova, urban blues, etc. - that the result will clearly be relevant. But new challenges met us when we set to work on the large and very different topic of folk music, which ranges from Scottish pipes and drums to Tibetan gya ling to American zydeco. The Danish edition of Dewey divides folk music into continents and countries but does not care about genres, instruments, or cross-cultural phenomena such as reggae, Nubian music, belly dance, flute, or banjo. We also had to find out the meaning behind words such as these: chouval bwa, biwa, tombak, qin, qawwali, bendir, and rolmo. Were they genres or instruments?

We decided to let our starting point be geographical division, continent by continent and country by country. Let's take a look at African Folk Music. We had not been working with that genre before and so we experienced an unknown realm with fantastic, mysterious and marvelous words. Some of them well-known but there were also a lot of new words that we have never heard of, or at least couldn't explain. Were they terms for instruments, genres, dances, or something quite different? Investigating the definitions of the terms did make it possible for us to group them. Some of them were specifically related to a particular country, as you see on the illustration below.

After making the above grouping we ran into problems with the aforementioned cross-cultural music. To solve these problems we invented two groupings that transverse each continent, for example "Criss-Crossing Africa" and "Folk Musical Instruments of Africa". The first containing afrobeat, Arabic Music, mandingo, and world music, for example; the second containing instruments such as balafon, kora, talking drum, and wassakumba. But what did the words mean and what did the instruments look and sound like?

To enrich the library users' experience, we had an idea: Why not make a small pop-up window with a dictionary explaining the unfamiliar words, and attach a relevant link to a Web site with more detailed information? So that is what we did. To visualize the geographic relations and to supplement the navigation facilities we have begun to add maps of the continents where each country has its own link beneath which lies a search query into the library database.

Figure 1:
Each country has its own musical terms.

Figure 2:
Criss-Crossing Africa and Folk Musical Instruments of Africa: By clicking on red red buttom you can open the dictionary with explanations an links to useful external Web sites.
Figure 3:
The dictionary is contructed as a separate pop-up window containing an alphabetical list of all the foreign terms.

Our Experience Helps Us Plant Future Seedlings
As we worked, we found out that each type of music has its own characteristics. For example, what we were able to use as a model for creating a structure and a grouping within jazz could not be used for the rock section, and what we did with the rhythm & blues section could not be used when we came to the section on world music.

Last summer when we completed the first greater group, jazz, we requested patrons to participate in a usability test. The result was overwhelmingly positive; the patrons felt that it was a tool they had been missing for a long time and they asked for new categories added to MusiCat. We have scheduled a professional usability testing company to test MusiCat in August 2002. Only by continuously and repeating usability tests you can know how to improve the system so it can live up to our and their expectations regarding these goals:

  • Easy accessibility
  • Easy navigation
  • Easy location of previously hidden information
  • Easy-to-use display of the contents of the database
  • A good association tool for inspiration, delight, and amusement
  • A tool for systematic navigation to the world of music

To date, we have completed jazz and folk music. Rock, blues, gospel, and easy listening are on their way. We have yet to begin classical music and special groups like national anthems, religious music, film music, signature tunes, new age, music for relaxation, gymnastics, and others. From the beginning, it has been very important to us that MusiCat should maintain and update itself after we have completed the construction of the various directories. This means that once constructed, the maintenance should preferably be kept to a minimum, and it means that the content should automatically be renewed each time new items are added to the database or old items are deleted from the database. To achieve that functionality, every search string has been constructed dynamically: That is to say each consists of elements such as keywords, subject headings, or other universal terms. We have deliberately chosen never to use the names of artists, composers, conductors or bands in order to avoid static search strings.

Until now we have been operating with the classic media: CDs, cassette tapes, records, and musical scores. However, videotapes, music videotapes, and DVDs are relevant, too. I think the best solution is to make a Web site especially for those media.

Our next plan is to get permission to add small sound bytes (20-30 seconds each) of a selection of characteristic genres and instruments by paying the necessary copyright agencies for the rights to use them. By clicking on an icon of a loudspeaker you will then be able to listen to a small sample of the desired music.

In our efforts to make our Web site and our catalog even more user-friendly through MusiCat, group by group of different styles and categories of music have found their places online at Silkeborg Public Library. I expect that these useful search trees will continue to grow and flourish.

For Further Exploration

Per Hofman Hansen is the Web site editor in chief and librarian at Silkeborg Public Library in Silkeborg, Denmark. He graduated from the Royal School of Library and Information Science in Copenhagen, then worked for a year at the Royal Library before he began working at Silkeborg. He has worked there as head librarian of the department of fiction and classed books for 25 years. For the last 12 years he has been a "censor" (Danish term for academic independent examiner) at the Royal School of Library and Information Science. His e-mail address is
(Note: The author would like to thank Karen and Tim Williams for their help translating this article into English).

To my homepage. . Published on the Internet January 1, 2003. Updated February 12, 2005.